* [dpdk-users] DPDK layer 2 forwarding without modification / disable offloading
@ 2021-07-13 11:29 dennis.wagner
0 siblings, 0 replies; only message in thread
From: dennis.wagner @ 2021-07-13 11:29 UTC (permalink / raw)
I am a student of computer science in Germany. Sorry, if this is a basic question.
I am trying to do some latency benchmarks with DPDK as a layer 2 relay. Here is what I did:
* my server is running Debian 10 and has 2 NICs Intel I210 Gigabit
* NICs are configured: dpdk-devbind.py --bind=uio_pci_generic 0000:02:00.0 / 0000:05:00.0
* examples/skeleton/basicfwd.c is used to relay traffic
* to measure the latency, traffic is recorded before and after the dpdk-box with tcpdump
* compare the 2 traces of a http download
I expected DPDK to pass all frames unmodified, but some frames after the handshake the traces start to diverge: the length of the frames change.
Could hardware offloading like TSO be the reason? How would I completely disable offloading?
Is this done in my DPDK application or with a tool like ethtool per interface?
Setting the offloads field in basicfwd.c (port_conf.txmode.offoads = 0; port_conf.rxmode.offoads = 0) did not work.
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2021-07-13 11:29 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-07-13 11:29 [dpdk-users] DPDK layer 2 forwarding without modification / disable offloading dennis.wagner
DPDK usage discussions
This inbox may be cloned and mirrored by anyone:
git clone --mirror http://inbox.dpdk.org/users/0 users/git/0.git
# If you have public-inbox 1.1+ installed, you may
# initialize and index your mirror using the following commands:
public-inbox-init -V2 users users/ http://inbox.dpdk.org/users \
Example config snippet for mirrors.
Newsgroup available over NNTP:
AGPL code for this site: git clone https://public-inbox.org/public-inbox.git